TM1 Rule BY REFERENCE instead of default BY VALUE
Posted: Thu Aug 11, 2011 8:36 pm
Consider
Actuals N
Scenarios C
Scenario1 N
Scenario2 N
...
Scenario99 N
We want to rule all scenarios to show actuals for the first 2 quarters
['Scenarios',{'Q1','Q2'}] = N: ['Actuals'] ;
FEEDERS ;
['Actuals','Q1'] => ['Scenarios','Q1'] ;
['Actuals','Q2'] => ['Scenarios','Q2'] ;
Now consider we have 99 scenarios: what these rules do feed 99 blank copies of Actuals for the first 2 quarters which are populated with values when any of the scenarios has q1 or q2 data accessed ( in a view or a TI etc).
These 99 copies take time on cube start up but they contain purely redundant information. In old style programming terms we can refer to a variable by a pointer BY REFERENCE or pass or copy the variable BY VALUE.
If we use BY REFERENCE we effectively inform the interpreter/compiler to look at a single copy of the Actuals data and there is no redundancy.
Forecasting is commonly where a known stream splits into several scenarios. Tm1, being a good forecasting tool, may have solved this problem.
I wonder is there a facility to refer to a single reference to actuals rather than making multiple copies. Is there plans to introduce something like this in the future or is the copy approach going to be acceptable now we have faster 64 bit machines with bags of memory.
Thanks
John
Actuals N
Scenarios C
Scenario1 N
Scenario2 N
...
Scenario99 N
We want to rule all scenarios to show actuals for the first 2 quarters
['Scenarios',{'Q1','Q2'}] = N: ['Actuals'] ;
FEEDERS ;
['Actuals','Q1'] => ['Scenarios','Q1'] ;
['Actuals','Q2'] => ['Scenarios','Q2'] ;
Now consider we have 99 scenarios: what these rules do feed 99 blank copies of Actuals for the first 2 quarters which are populated with values when any of the scenarios has q1 or q2 data accessed ( in a view or a TI etc).
These 99 copies take time on cube start up but they contain purely redundant information. In old style programming terms we can refer to a variable by a pointer BY REFERENCE or pass or copy the variable BY VALUE.
If we use BY REFERENCE we effectively inform the interpreter/compiler to look at a single copy of the Actuals data and there is no redundancy.
Forecasting is commonly where a known stream splits into several scenarios. Tm1, being a good forecasting tool, may have solved this problem.
I wonder is there a facility to refer to a single reference to actuals rather than making multiple copies. Is there plans to introduce something like this in the future or is the copy approach going to be acceptable now we have faster 64 bit machines with bags of memory.
Thanks
John